App note: Layered isolation with high power digital isolators

Posted on Sunday, September 23rd, 2012 in app notes by DP

Sequentially connecting optocouplers for better isolation slows things way down, digital isolators don’t suffer as much. The advantages of digital isolators over opto-couplers [PDF!] in layered isolation:

Creating double insulation barriers with sequential optocouplers is problematic because the data integrity is poor, and there is no compact inexpensive way to provide power to the interface between the barriers. With the advent of high performance digital isolators like iCoupler, and integrated power in isoPower devices, creating high voltage isolation
barriers by layering isolators is now a viable solution.

This entry was posted on Sunday, September 23rd, 2012 at 9:00 pm and is filed under app notes. You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.

7 Responses to “App note: Layered isolation with high power digital isolators”

  1. Guan Yang says:

    Could you explain briefly what digital isolates are? The Wikipedia article is just full of marketing speak.

    • bearmos says:

      They’re similar to opto-isolators – except they don’t use optics ;-)

      Optical isolators use an LED to signal a photo-transistor. The problem with opto-isolators is that they can take a fair amount of current to turn on the LED, they’re slow because of a slow response from the photo-transistor), and they *can* have a limited life (this is according to some of the digital isolator datasheets, since the LED has a finite lifetime – still many many years).

      Digital isolators accomplish isolation without the LED and photo-transistor, usually by modulating the signal, transmitting via RF, and demodulating. Others will use transformers to do the job (like the iCoupler mentioned in this app note), there are probably other technologies out there as well.

      Digital isolators can usually operate at faster speeds, sometimes higher temps, don’t require as much current to drive the lines (since there’s no LED to drive), and are suppose to last longer. However, this comes at a price. Most times, digital isolators require two separate power supplies (one to power each half of the isolator) – there are exceptions to this, like the isoPower form analog devices, which is basically just an isolated dc-dc converted on the chip. There is also some concern on the DP forums that digital isolators might induce some noise if used on sensitive analog boards by inducing noise into the power supply lines (fast transients from switching at RF), but this hasn’t been fully empirically investigated yet.

      Wikipedia’s opto-isolator page gives some more detail:

      Here are some assorted datasheets and product pages:

    • Ian says:

      The Isolators (both kinds) create a connection between two circuits without actually “touching” each other’s power supply. Optoisolators do this by shining a LED into a photo resistor which acts as a on/off switch. This is pretty slow, so digital isolators use magnetic or other techniques to control switches.

  2. MattB says:

    This is pretty cool. I’ve had very good experiences using the AD iCoupler range for RS485 isolation in several designs now. The isoPower options are a bit pricey but they can really cut down on part count and design time (eg. I put together an isolated 485 to USB adapter for in house testing with just two bits of silicon and a few passives).

  3. Another type is NVE’s range of giant magnetoresistive isolators:

    These use the GMR effect to transmit transitions across an insulating barrier. Speeds and delay times are pretty incredible – 100MBps, and 10ns delays. I’ve used the IL710 in a MOSFET driver before.

    The only downside is that the output state is undefined on power-up, so you need to take account of that. Caused some headaches before!

  4. bearmos says:

    “spintronic GMR technology” . . . sounds fancy :) Definitely a new one for me.

    Great content on the site, btw – eventually I might be looking into doing some SS etching for SMD stencils. I was thinking about trying out the salt solution (similar to what you had done), but I’m wonder if it’s worth the bother now. It sounds like standard photo resist works well as a steal etch resist. I wonder if there’s a grade of SS that electro-etches well with salt water but still has the desirable physical properties for an SMD stencil (I’d rather not have another chemistry experiment in my basement!).

    With DP’s current kick on nixie tubes, Ian et al might like your home-brew nixies (for lack of a better term):

    • Thanks – actually, the electrolytic etching has some problems when you try and etch large flat things. I tried doing a thing similar to a stencil (thin sheet with a design all over) and the bits near the edge etched far more (and hence undercut) than the bits near the middle. Maybe I needed a bigger counter electrode, farther away, but I was pretty careful to rotate it to get things uniform.

      Better is probably a purely chemical etch, if you can find something that’ll do stainless. Actually, I never
      quite saw why the stencil had to be stainless – sure, it’s better for wear resistance if you’re doing a lot, but you can get resist-coated brass shim (e.g. which would etch nicely with ferric chloride.

      Some stainless seems to etch a lot better than others – the kitchen knife I tried went great, but a piece of 304 sheet did not (very rough etch). Possibly higher carbon content is better.

      If you can get near a laser cutter, plastic film stencils work quite well.

      Re: nixie tubes, you might like to see and The guy is really something!

Leave a Reply

Notify me of followup comments via e-mail. You can also subscribe without commenting.

Recent Comments

  • Pavel: dangerousprototypes the best!
  • Leon: Too late?
  • Drone: First of all, let me post a reference about what Action Potential is in terms of Physiology (since DP isn't adding much value these days...
  • Daniel: Isn't it funny how free Sunday PCBs is always posted on Sunday night or Monday (in most parts of the world)?
  • hli: Sunday++